Commentary: Using innovative inequality measures in epidemiology
Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Canada.
*Corresponding author. Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Canada. E-mail: sam.harper{at}mcgill.ca
Accepted 6 June 2007
Zhang and Wang1 use the concentration index (CI) to examine changes in income-related inequalities in overweight among US adolescents (aged 10–18) between 1971 and 2000. Their findings were heterogeneous and do not suggest any simple interpretation. They concluded that the expected inverse associations between income and overweight were not always present and differed across race/ethnic groups and over time. Between 1971 and 2000, the extent of income-related inequality in overweight generally diminished, mainly because of growing overweight among more socially advantaged adolescents. They recommend that interventions should not focus only on socially disadvantaged groups but instead target the whole population because there are many overweight adolescents in every social group.
There are several themes that emerge from Zhang and Wang's work that are relevant to understanding and measuring how social factors affect health. The heterogeneity of their findings implies specificity of mechanisms linking social processes with health processes, in this case overweight. There is no general way that income-related inequality affects inequality in overweight among all adolescents because the associations differ across race/ethnic groups, gender and time. This suggests that other social processes not simply related to economic inequality are acting to produce the differences observed across race/ethnic groups. In other words, income does not tell us as much as we might like to know about the social distribution of overweight and it tells us even less about its change over time. The declining inequality in overweight is largely driven by social forces that are immune to whatever higher income usually offers as protection from ill health. This can be seen just as easily among US adults. Figure 1 shows that while there are measurable differences in average body mass index (BMI) by education level, the within-group distribution of BMI far outweighs between-group differences. Socioeconomic position does not appear to be a particularly good indicator of the health risks associated with high BMI.
|
It is important in epidemiology to be clear on what it is that needs to be explained. Studies such as this are one important tool in monitoring changes in the extent of social inequalities. Once we understand what has changed, then we may be in a better position to understand what caused those changes. For example, in recent work2 we showed how the black–white life expectancy gap in the US declined markedly over the last decade to reach an historic low. By decomposing the age- and cause-specific structure of those changes, we showed that
70% of the declining gap was due to relative improvements in just 3 or 4 causes of death—HIV, homicide, unintentional injuries and, among women, heart disease. In other words, the processes (like the ones implied by Zhang and Wang's findings) were specific, not general, which suggests particular mechanisms of action linking social processes with health processes. We applaud the use of innovative measures of health inequality in epidemiology.3 This point has been made several times in the past4,5 and such measures are increasingly appearing in the epidemiologic literature.6–10 Zhang and Wang use the relative CI—a tool for measuring inequality with a long history in economics11,12 but relatively recently developed and applied to health.13,14 The CI belongs to a general class of indicators sometimes called measures of ‘disproportionality’15 that includes the well-known Gini coefficient, and less well-known measures of entropy such as the ‘Mean Log Deviation’ and ‘Theil Index’ (among others). These measures all express inequality as a function of shares of some health outcome (in this case overweight) compared with shares of the population, in some cases, ordered by social group (in this case income).
Given the suite of potential tools for measuring health inequality, the choice of a particular measure invariably entails trading off advantages and disadvantages that are sometimes not obvious. Zhang and Wang point out that one limitation of the CI is that it may only be used with ordinal measures of social group status. But this only limits it as a generalized measure of social inequalities in health. Given their interest in measuring socioeconomic inequalities, this is hardly a limitation. Nevertheless, the CI does have other characteristics that could be seen as limitations, namely: (i) it is insensitive to changes in socioeconomic position that do not affect socioeconomic ranking (e.g. an increase in the absolute level of deprivation of the poor that does not affect their socioeconomic rank will have no effect on the CI16); (ii) when the health outcome of interest is a dichotomous state (i.e. obese/not obese) the range of the CI is limited by the overall prevalence of the outcome,17 which is particularly important for comparing inequalities over time periods when prevalence changes substantially and (iii) the CI uses the population average rate of health/disease as the reference point for measuring inequalities. This limitation is important for health inequality assessments because declines in inequality as measured by the CI may be generated by worsening health among those with initially better health. This applies to other measures of health inequality as well, but brings up a more fundamental question, and that is whether the measurement of health inequalities can and should be separated from the measurement of average health itself. If health inequality declines because the health of a healthier, more advantaged social group declines, it would seem inconsistent with more general principles of public health to suggest the overall health situation of the population has improved. But this is precisely what the CI would indicate. It may be worth thinking about measures of inequality that are sensitive to both the absolute level of health and its social distribution.
Despite its limitations, Zhang and Wang note an advantage of the CI, which is that it takes into account the entire range of socioeconomic groups in the population. This is indeed true, but the CI has other potential advantages as a tool for monitoring inequality. First, while the CI is commonly used as a measure of relative inequality, it may also be used to measure health inequalities on an absolute scale.3 This is important because relative and absolute inequality often diverge over time in the context of strong secular trends. For example, Figure 2 shows that while the relative CI for educational inequality in current smoking in the US increased by roughly 50% from 1988 to 2003, the absolute CI was virtually unchanged because smoking declined in all groups. Second, the CI is a ‘decomposable’ index, both in terms of quantifying how much of the change in inequality is due to changes in the population distribution of socioeconomic position vs changes in health, and because it allows quantification of the contribution of changes in the social determinants of health inequality to changes in health inequality.3 In the first case, it seems important that an inequality measure is sensitive to shifts in the population distribution of exposure. For instance, given strong secular trends in education, the proportion of the population exposed to the lowest levels of education has decreased. If such ‘upstream’ changes due to shifts in educational opportunity per se have occurred then it seems important to account for that in assessments of changes in health inequality over time. The other potentially useful feature of measures like the CI is that decomposition allows quantification of how much of the socioeconomic inequality in health is due to socioeconomic inequalities in health determinants.18 For instance, Hosseinpoor and colleagues10 recently showed that while 36% of income-related inequality in infant mortality in Iran was due to economic inequality per se, 64% was due to the fact that important risk factors for infant mortality such as maternal illiteracy and access to a hygienic toilet were strongly correlated with income. Thus, decomposition is important because it links monitoring health inequality with understanding its determinants. Third, the CI may be modified to be more or less sensitive to the health of certain socioeconomic groups though the use of a parameter indicating the level of ‘aversion’ to inequality.19 This is often not made explicit, but the standard CI (as used by Zhang and Wang) essentially weights the health of the poorest individual in the population by a factor of 2 and this weighting factor declines as socioeconomic rank increases. This may be a reasonable scale for weighting the health of different socioeconomic groups, but it is nevertheless arbitrary. There is no reason why, for any given outcome, we could not raise (or lower) the weight given to the health of the poor. This is an important but undervalued reason for using the CI because the aversion parameter could be calibrated to be consistent with the public and/or policymakers’ attitudes toward the extent of inequality. Research aimed at eliciting such preferences would seem worthwhile.
|
Finally, Zhang and Wang's article provides an opportunity to reflect on an important connection between inequality measurement and ethical values20 that is not often made explicit. Statistical aspects of inequality measures, such as whether they weight by population size, the reference point from which they measure differences in health, and whether they are measured on the absolute or relative scale, also imply judgements about what aspects of inequality are thought to be important. For example, weighting social groups by population size when measuring inequality implies giving the health of each individual equal consideration, whereas ignoring population size implies giving the health of social groups equal consideration. One could make a case for either of these two perspectives, but it should be clear that it is not simply a scientific argument but also a moral argument about how much to value the lives of individuals. Similarly, one of the major reasons for using the CI to measure socioeconomic inequalities in health is because it gives added weight to the health of the least well-off. Explicitly incorporating such a priori considerations when measuring health inequalities both helps to make clear the motivation for choosing particular measures of inequality, and serves to strengthen the links between the moral intuitions that underlay our concern with social inequalities in health and the tools we use to quantify them.
 |
References |
|---|
 Top References |
|---|
1 Zhang Q, Wang Y. Using concentration index to study changes in socio-economic inequality of overweight among US adolescents between 1971 and 2002. Int J Epidemiol (2007) 36:916–925.
2 Harper S, Lynch J, Burris S, Davey Smith G. Trends in the black-white life expectancy gap in the United States, 1983–2003. JAMA (2007) 297:1224–32.
3 Harper S, Lynch J. Methods for Measuring Cancer Disparities: A Review Using Data Relevant to Healthy People 2010 Cancer-Related Objectives (2006) Washington DC: National Cancer Institute. Retrieved June 6, 2007 from http://seer.cancer.gov/publications/disparities/.
4 Wagstaff A, Paci P, van Doorslaer E. On the measurement of inequalities in health. Soc Sci Med (1991) 33:545–57.[CrossRef][ISI][Medline]
5 Mackenbach JP, Kunst AE. Measuring the magnitude of socio-economic inequalities in health: an overview of available measures illustrated with two examples from Europe. Soc Sci Med (1997) 44:757–71.[CrossRef][ISI][Medline]
6 Zhang Q, Wang Y. Socioeconomic inequality of obesity in the United States: do gender, age, and ethnicity matter? Soc Sci Med (2004) 58:1171–80.[CrossRef][ISI][Medline]
7 Gravelle H, Sutton M. Income related inequalities in self assessed health in britain: 1979-1995. J Epidemiol Commun Health (2003) 57:125–29.
8 Clarke P, Smith L, Jenkinson C. Comparing health inequalities among men aged 18-65 years in Australia and England using the SF-36. Aust N Z J Public Health (2002) 26:136–43.[Medline]
9 Humphries KH, Van Doorslaer E. Income-related health inequality in Canada. Soc Sci Med (2000) 50:663–71.[CrossRef][ISI][Medline]
10 Hosseinpoor AR, Van Doorslaer E, Speybroeck N, et al. Decomposing socioeconomic inequality in infant mortality in Iran. Int J Epidemiol (2006) 35:1211–19.
11 Kakwani NC. Applications of Lorenz curves in economic analysis. Econometrica (1977) 45:719–28.[CrossRef][ISI]
12 Mahalanobis PC. A method of fractile graphical analysis. Econometrica (1960) 28:325–51.[CrossRef][ISI]
13 Wagstaff A, Van Doorslaer E, Paci P. Equity in the finance and delivery of health care: some tentative cross-country comparisons. Oxf Rev Econ Policy (1989) 5:89–112.
14 Kakwani N, Wagstaff A, van Doorslaer E. Socioeconomic inequalities in health: measurement, computation, and statistical inference. J Econometr (1997) 77:87–103.[CrossRef]
15 Firebaugh G. The New Geography of Global Income Inequality (2003) Cambridge, MA: Harvard University Press.
16 Erreygers G. Beyond the health concentration index: an Atkinson alternative for the measurement of the socioeconomic inequality in health. In: UNU-WIDER Conference on Advancing Health Equity (2006) Helsinki, Finland: United Nations University, World Institute for Development Economics Research. Retrieved June 6, 2007 from http://www.wider.unu.edu/conference/conference-2006-2/conference-2006-2-papers/erreygers-170806.pdf.
17 Wagstaff A. The bounds of the concentration index when the variable of interest in binary, with an application to immunization inequality. Health Econ (2005) 14:429–32.[CrossRef][ISI][Medline]
18 Wagstaff A, van Doorslaer E, Watanabe N. On decomposing the causes of health sector inequalities with an application to malnutrition inequalities in Vietnam. J Econom (2003) 112:207–23.[CrossRef][ISI]
19 Wagstaff A. Inequality aversion, health inequalities and health achievement. J Health Econ (2002) 21:627–41.[CrossRef][ISI][Medline]
20 Temkin LS. Inequality (1993) New York: Oxford University Press.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||